DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
The present application, filed on February 19, 2025, is accepted.
Claims 1 – 20 are being considered on the merits.
Drawings
The drawings, filed on February 19, 2025, are accepted.
Specification
The specification, filed on February 19, 2025, is accepted.
Double Patenting
No rejection warranted at application’s initial filling time of filling for a patent.
Claim 6 is objected to because of the following informalities:
In claims 6, it recites “QS” test. The first instance of an abbreviation needs to be full expanded. Appropriate correction is required.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1 – 5, 7 – 16, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over US 20180324199 A1 to Crotinger et al., (hereinafter, “Crotinger”) in view of US 20210089927 A9 to Ryan et al., (hereinafter, “Ryan”).
Regarding claim 1, Crotinger teaches a computer-implemented method for detecting operational anomalies in a multi-domain communication network, the method comprising: obtaining a plurality of time series of performance data from multiple domains of the communication network; [Crotinger, para. 7 discloses the underlying statistical model may take into account trends, seasonality, and/or other patterns that may be relevant to predicting expected subsequent time-series data values. As additional time-series data is received, it is compared with the underlying statistical model to identify statistical outliers in the time-series data. The statistical outliers are attributed an anomalous score, indicating an attributed magnitude of anomaly for the statistical outlier. The anomalous scores for the statistical outliers are then provided via the system, enabling anomaly reporting and/or remedial action.] determining one or more models of non-anomalous network behavior based on the plurality of time series; [Crotinger, para. 34 discloses accumulated time-series data may be analyzed to classify the time-series data by one or more qualitative classifications. The classifications may be used to generate a statistical model of the time-series data, which may help to identify anomalies (e.g., outlier data that falls outside of upper and/or lower bounds of the statistical model). Anomalous scores for the time-series data may be generated based at least in part upon an accumulation of anomalies over a certain dynamically-adjustable window of time.] classifying the respective time series into a plurality of types based on the presence or absence of at least two types of components in the respective time series; [Crotinger, para. 74 discloses time-series data that appears to adhere to a stationary process and is, thus, classified with a quasi-normal classification, in accordance with an embodiment. A stationary process is a stochastic process whose joint probability distribution does not change when shifted in time. Consequently, parameters such as mean and variance, if they are present, also do not change over time. FIG. 5 is a plot 500 of the percentage of free memory for a server that does not have a strongly seasonal (hourly, daily or weekly periodic) component, along with a plot of the estimated probability density for this signal. The data of FIG. 5 have a mean of 51.7 and a standard deviation of 5.7. For this kind of signal, outliers are easily identified using variance based metrics or derivations, such as by characterizing measurements outside of μ±3σ; (e.g., outside of the range 34.5-68.9%) as outliers. In order to keep anomalies from having a large effect on the future control bounds, a robust estimation of the standard deviation may be used instead of the sample standard deviation itself.], but Crotinger does not teach detecting for operational anomalies, based on the one or more models and the classified types, in the plurality of time series or in further performance data obtained from the multiple domains of the communication network.
However, Ryan does teach detecting for operational anomalies, based on the one or more models and the classified types, in the plurality of time series or in further performance data obtained from the multiple domains of the communication network. [Ryan, para. 17 discloses a system for detecting outliers of network data. The system comprises one or more processors and a memory in communication with the one or more processors. The memory is configured to store instructions for detecting outliers of network data. The instructions, when executed, cause the one or more processors to obtain network data from a network to be monitored, and create a window from the obtained network data. The instructions further cause the one or more processors to detect outliers of the obtained data with respect to the window using an unsupervised learning process including one or more of a Generalized Adversarial Network (GAN) learning technique and a Bidirectional GAN (BiGAN) learning technique, whereby the unsupervised process does not require human intervention.]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Ryan’s system with Crotinger’s system, with a motivation for pattern detection, and more particularly relates to pattern detection in time-series data. Conventional systems may be used to detect patterns, but typically do not detect patterns in data obtained specifically from a time-series. Pattern detection in a time-series, according to the present disclosure, may be used in the field of Machine Learning (ML) for networking applications, telecommunications, as well as many other applications. [Ryan, para. 61]
As per claim 2, modified Crotinger teaches the method of claim 1, further comprising, based on detecting a plurality of operational anomalies in the further performance data, determining an order of importance of the detected operational anomalies based on respective deviations from corresponding non-anomalous network behavior. [Crotinger, para. 7 discloses the underlying statistical model may take into account trends, seasonality, and/or other patterns that may be relevant to predicting expected subsequent time-series data values. As additional time-series data is received, it is compared with the underlying statistical model to identify statistical outliers in the time-series data. The statistical outliers are attributed an anomalous score, indicating an attributed magnitude of anomaly for the statistical outlier. The anomalous scores for the statistical outliers are then provided via the system, enabling anomaly reporting and/or remedial action.]
As per claim 3, modified Crotinger teaches the method of claim 2, further comprising one or more of the following: in response to one or more detected anomaly determined to be most important, initiating one or more corrective actions in a plurality of the domains of the communication network; [Crotinger, para. 43 discloses the MID server 126 (or a separate time-series database server) may classify time-series data based upon data patterns/trends of historical time-series data. The MID server 126 may generate and provide a model for the time-series data, which may be used to determine an expected range for subsequent time-series data or otherwise used to detect that a sequence of time-series data points are highly unlikely. The MID server 126 may accumulate time-series data outside of the expected range and generate an anomalous score for the time-series data, which may be provided to an active Instance 130 for subsequent remedial action/reporting.] and in response to one or more detected anomalies determined to be less important, refraining from initiating one or more further corrective actions in one or more domains of the communication network. [Crotinger, para. 35 discloses the time window may be adjusted to be relatively longer when a probability of the time-series data falling in a normal range is low. This may extend the monitoring window, such that even slight anomalies are monitored, when there is a low probability that the time-series data falls into a normal range. In such embodiments, the time window may also be adjusted to be relatively shorter when the probability of the time-series data falling in a normal range is high. Additional details regarding time-series data anomaly detection techniques are provided below with reference to FIGS. 1-16.]
As per claim 4, modified Crotinger teaches the method of claim 1, wherein one or more of the following applies: each time series comprises data samples from one of the following in a single domain: a network element, or an interface between network elements; [Crotinger, para. 28 discloses CIs may include information related to a physical entity (e.g., hardware), a logical entity (e.g., version, instance of a database), a conceptual entity (e.g., service), and the like associated with a respective device. Given the wide variety of CIs associated with various devices within this type of complex IT environment, configuration item (CI) discovery executed on a given infrastructure is used to track the CIs of the devices that are present on the connected IT environment. That is, CI discovery is the process of finding configuration items, such as hardware, software, documentation, location, and other information related to the devices connected to a given network, such as an enterprise's network.] and the operational anomalies are detected in a plurality of time series collected from a plurality of domains. [Crotinger, para. 30 discloses for a given computer system, data regarding performance of computer resources (e.g., central processing unit, disk, memory transaction count), instance cache size, code performance data (e.g., business rules count, cache size), large record counts (e.g., open incidents, tests), and the like may be stored at various times in a database (e.g., time-series database) accessible to a server system. As the amount of stored data increase, along with date/time stamps regarding when the data was acquired, it becomes increasingly difficult to analyze the stored data to identify subsets of the data that may be relevant to a particular issue or problem.]
As per claim 5, modified Crotinger teaches the method of claim 1, wherein classifying the respective time series based on the presence or absence of at least two types of components comprises: detecting whether each of the time series includes a seasonal component or a non-constant trend component; [Crotinger, para. 61 discloses Pattern detection in a time-series, according to the present disclosure, may be used in the field of Machine Learning (ML) for networking applications, telecommunications, as well as many other applications. For example, in the field of networking applications, pattern detection can be used in the following use cases: for forecasting threshold crossings, for forecasting alarms, for forecasting quality-of-experience (QoE), for network anomaly detection, among others.] classifying a time series as a first type when the time series includes a seasonal component; [Crotinger, para. 133 discloses preprocessing, a lagging step is employed that slides a window with certain size over the time-series. In order to capture the seasonality correlations, sizes of slides are chosen equal to human behavior activities. For instance, the window sizes could include one day worth of samples, one week worth of samples, one month worth of samples, or samples over any other suitable time period corresponding to the cycles of the signal.] classifying a time series as a second type when the time series includes a non-constant trend component but does not include a seasonal component; [Crotinger, para. 77 discloses the graph 10 of the network traffic volume also illustrates examples of anomalies in the data. Pattern detection is trained with historical data and anomalies can be identified and labeled. For example, windows 12 are labeled with “Y” to indicate the existence of an anomaly and windows 14 are labeled “N” to indicate an absence (or non-existence) of an anomaly. Multiple anomaly types can be encoded by using more than a binary classifier of “Y” and “N.” In some cases, multiple anomaly types can be detected in the same windows 12, 14 to indicate other types of anomalies or other patterns.] and classifying a time series as a third type when the time series includes neither a non-constant trend component nor a seasonal component. [Crotinger, para. 181 discloses the present disclosure is further directed to embodiments for utilizing machine learning processes to detect outliers or network anomalies in network time-series data, particularly machine learning processes that are executed in an unsupervised manner. One of the current problems that exists with many types of Machine Learning (ML) systems is an issue referred to as labelling. In a supervised ML scenario, labels will be provided to the data to train the machines or machine-like models, particularly for training to detect certain patterns in a time-series.]
As per claim 7, modified Crotinger teaches the method of claim 5, wherein detecting for operational anomalies in the plurality of time series comprises: decomposing each time series classified as the first type into a seasonal component, a non-constant trend component, and a noise component; [Crotinger, para. 77 discloses the trend plot 704 reflects the long-term progression of the time-series data (e.g., an increasing and/or decreasing direction in the time-series data). Seasonal plot 706 represents the seasonal variation patterns (e.g., patterns at particular fixed and/or known time periods). Residuals plot 708 represents the residuals or remainder of the time-series data after the trend and seasonality is removed. Plot 710 illustrates the trend data added to the residual data. A projection of the expected behavior for this time-series is obtained by smoothing the trend and seasonality components of the STL decomposition. Upper and lower control bands can then be computed by offsetting by the standard deviation of the residual from the STL decomposition 700. In some embodiments, the upper and lower control bounds may be computed using a robust estimate, as discussed above.] and calculating upper and lower bounds applicable to all time series classified as the first type; [Crotinger, para. 79 discloses Data classified as trendy may be handled similarly to data classified as seasonal, except that only the trend is extracted. The trend may be extracted using a linear regression, but may also be done with a more robust technique, such as a Theil-Sen estimator. The standard deviation of the residual distribution is calculated, possibly using a robust estimate, and then used to compute control bounds as offsets from the projected trend.] and detecting for operational anomalies in each time series classified as the first type based on comparing one of the following to the upper and lower bounds: respective non-constant trend components, and respective noise components. [Crotinger, para. 34 discloses accumulated time-series data may be analyzed to classify the time-series data by one or more qualitative classifications. The classifications may be used to generate a statistical model of the time-series data, which may help to identify anomalies (e.g., outlier data that falls outside of upper and/or lower bounds of the statistical model). Anomalous scores for the time-series data may be generated based at least in part upon an accumulation of anomalies over a certain dynamically-adjustable window of time.]
As per claim 8, modified Crotinger teaches the method of claim 5, wherein: each time series classified as the third type includes a noise component; [Crotinger, para. If the data appears to be the accumulation (sum) of a low-noise signal, the historical data may be classified as accumulating data. FIG. 9 is an example of time-series data 900 showing the total number of jobs on a particular server. This value is expected to increase at a steady rate and can be well fit with linear regression.] and detecting for operational anomalies in the plurality of time series comprises: decomposing each time series classified as the second type into a non-constant trend component and a noise component; [Crotinger, para. 82 discloses the “Accumulator” classification may also be fit with a linear approximation, but the residual may typically not be random. Thus, control bounds may not be an appropriate outlier detection mechanism. These time series may be transformed by differencing (e.g., monitoring the difference between successive observations instead of monitoring the metric time series itself). Accumulator-classified data may often be a near-constant series that in turn is monitored for change detection, rather than for data points that fall outside of the control limits.] and detecting for operational anomalies in each time series classified as the second type or the third type based on the respective noise components. [Crotinger, para. 83 discloses If the data appears to adhere to a set of discrete values, the historical data may be classified as multinomial or categorical. For example, FIG. 10 is an example of time-series data 1000 that shows the average number of semaphores in use on a particular server along with its frequency distribution. Because this time-series data includes a set of discrete values, the time-series data is classified with a multinomial or categorical classification, in accordance with an embodiment.]
As per claim 9, modified Crotinger teaches the method of claim 8, wherein: each noise component includes a series of tuples, with each tuple including a data value and a corresponding time instant; [Crotinger, para. 88 discloses characterizing classified time-series data, in accordance with an embodiment. This characterization may be used to generate the underlying statistical model 337 of FIG. 3. The process 1300 begins by receiving the time-series data classification derived from block 408 of the process 400 of FIG. 4 (block 1302). If the classification is quasi-normal (decision block 1304), the time-series data may be characterized using a mean and standard deviation of the time-series data (block 1306).] and detecting for operational anomalies in each time series classified as the second type or the third type comprises: rescaling the data values or the time instants comprising the tuples of the noise component; [Crotinger, para. 89 discloses If the classification is seasonal (decision block 1308), it may be beneficial to extract the seasonality from the time-series data. Extracting seasonality enables projection of dynamic control bounds which may provide an accurate statistical forecast of where future data is expected to lie. Thus, the extraction of seasonality enables more accurate identification of points that are far from where they are expected to lie. Ignoring these seasonal trends may lead to false positives, when a strong daily or weekly spike might register as an anomaly if statistics had been computed without adjusting for seasonality. Similarly, false negatives may be possible, as a data point may be outside the dynamic (seasonally-adjusted) upper control bound in the trough of a seasonal pattern but be inside a static control bound computed without first extracting the seasonal component.] and detecting for operational anomalies based on arranging the tuples into a plurality of clusters, including a cluster of non-outliers and at least one cluster of outliers. [Crotinger, para. 97 discloses the anomaly detector 332 may monitor the stream 313 of time-series/metric data. The anomaly detector 332 identifies statistical outliers of a current stream 313 of the time-series/metric data, by performing a statistical analysis on the stream 313 of the time-series/metric data based at least in part upon the underlying statistical model 337 constructed by the time-series analyzer 330. FIG. 14 illustrates an example plot 1400 illustrating the relationship of the features of this comparison. Specifically, the anomaly detector 332 compares the actual observed data (e.g., stream 313) with the expected data 1402 (e.g., as determined based upon the underlying statistical model 337). Upper bounds 1404 and/or lower bounds 1406 may also be used to determine the outliers. For example, data may be determined to be an outlier if it falls outside the range of the upper bounds 1404 to the lower bounds 1406. For example, outliers 1408 are identified in the plot 1400, because they fall below the lower bounds 1406.]
Regarding claim 10, modified Crotinger teaches the method of claim 1, but Crotinger does not teach wherein determining one or more models of non-anomalous network behavior based on the plurality of time series comprises training one or more machine learning, ML, models based on the plurality of time series using L1 regularization.
However, Ryan does teach wherein determining one or more models of non-anomalous network behavior based on the plurality of time series comprises training one or more machine learning, ML, models based on the plurality of time series using L1 regularization. [Ryan, para. 138 discloses Providing models with high accuracy in detecting anomalies is generally a complex Nondeterministic Polynomial (NP) hard empirical process. Models have to be extensively replaced by other algorithms and optimized to avoid under-fitting when the input evolves to a more complex and heterogeneous data. Nevertheless, however, as the data grows with time, the models tend to become more complex and eventually over-fit the data. Techniques to avoid the expansion of data to an explosion point are regularization techniques such as removing old instances or assigning a higher weight to the recent inputs, randomly setting neural network weights to 0 during training, and limiting the magnitude of the weights during training.]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Ryan’s system with Crotinger’s system, with a motivation for pattern detection, and more particularly relates to pattern detection in time-series data. Conventional systems may be used to detect patterns, but typically do not detect patterns in data obtained specifically from a time-series. Pattern detection in a time-series, according to the present disclosure, may be used in the field of Machine Learning (ML) for networking applications, telecommunications, as well as many other applications. [Ryan, para. 61]
Regarding claim 11, modified Crotinger teaches the method of claim 10, but Crotinger does not teach wherein: each ML model comprises a neural network, NN, having a plurality of weights; and training the one or more ML models using L1 regularization comprises minimizing, for each ML model, a loss function of the NN weights and of a norm of the NN weights.
However, Ryan does teach wherein: each ML model comprises a neural network, NN, having a plurality of weights; [Ryan, para. 13 discloses The one-dimensional or multi-dimensional windows are created either independently or jointly with the time-series. The method also includes training a Deep Neural Network (DNN) with the one-dimensional or multi-dimensional windows utilizing historical and/or simulated data to provide a neural network model. The method further includes processing ongoing data from a network with the neural network model to detect one or more patterns of a particular category in the ongoing data and localizing the one or more patterns in time.] and training the one or more ML models using L1 regularization comprises minimizing, for each ML model, a loss function of the NN weights and of a norm of the NN weights. [Ryan, para. 119 – 124 discloses The preparation step is selected during the training of the machine learning algorithm. The multi-dimensional scheme may be produced in many ways, such as by: [0120] stacking transformed streams without modifications; selecting one transformed stream and return it; obtaining a weighted sum of transformed streams; multiplying stacked streams by the matrix (multidimensional weighted sum); and passing stacked streams through a non-linear function (e.g., neural network).]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Ryan’s system with Crotinger’s system, with a motivation for pattern detection, and more particularly relates to pattern detection in time-series data. Conventional systems may be used to detect patterns, but typically do not detect patterns in data obtained specifically from a time-series. Pattern detection in a time-series, according to the present disclosure, may be used in the field of Machine Learning (ML) for networking applications, telecommunications, as well as many other applications. [Ryan, para. 61]
Regarding claim 12, modified Crotinger teaches the method of claim 10, but Crotinger does not teach wherein detecting for operational anomalies based on the one or more models comprises, using the one or more trained ML models, predicting non-anomalous network behavior in one or more of the following: a second portion of the plurality of time series, different than the first portion used to train the one or more ML models; and the further performance data obtained from the multiple domains of the communication network.
However, Ryan does teach wherein detecting for operational anomalies based on the one or more models comprises, using the one or more trained ML models, predicting non-anomalous network behavior in one or more of the following: a second portion of the plurality of time series, different than the first portion used to train the one or more ML models; [Ryan, para. 13 discloses The one-dimensional or multi-dimensional windows are created either independently or jointly with the time-series. The method also includes training a Deep Neural Network (DNN) with the one-dimensional or multi-dimensional windows utilizing historical and/or simulated data to provide a neural network model. The method further includes processing ongoing data from a network with the neural network model to detect one or more patterns of a particular category in the ongoing data and localizing the one or more patterns in time.] and the further performance data obtained from the multiple domains of the communication network. [Ryan, para. 138 discloses Providing models with high accuracy in detecting anomalies is generally a complex Nondeterministic Polynomial (NP) hard empirical process. Models have to be extensively replaced by other algorithms and optimized to avoid under-fitting when the input evolves to a more complex and heterogeneous data. Nevertheless, however, as the data grows with time, the models tend to become more complex and eventually over-fit the data. Techniques to avoid the expansion of data to an explosion point are regularization techniques such as removing old instances or assigning a higher weight to the recent inputs, randomly setting neural network weights to 0 during training, and limiting the magnitude of the weights during training.]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Ryan’s system with Crotinger’s system, with a motivation for pattern detection, and more particularly relates to pattern detection in time-series data. Conventional systems may be used to detect patterns, but typically do not detect patterns in data obtained specifically from a time-series. Pattern detection in a time-series, according to the present disclosure, may be used in the field of Machine Learning (ML) for networking applications, telecommunications, as well as many other applications. [Ryan, para. 61]
Regarding claim 13, modified Crotinger teaches the method of claim 12, but Crotinger does not teach wherein detecting for operational anomalies is based on the non-anomalous network behavior predicted using the one or more trained ML models.
However, Ryan does teach wherein detecting for operational anomalies is based on the non-anomalous network behavior predicted using the one or more trained ML models. [Ryan, para. 138 discloses Providing models with high accuracy in detecting anomalies is generally a complex Nondeterministic Polynomial (NP) hard empirical process. Models have to be extensively replaced by other algorithms and optimized to avoid under-fitting when the input evolves to a more complex and heterogeneous data. Nevertheless, however, as the data grows with time, the models tend to become more complex and eventually over-fit the data. Techniques to avoid the expansion of data to an explosion point are regularization techniques such as removing old instances or assigning a higher weight to the recent inputs, randomly setting neural network weights to 0 during training, and limiting the magnitude of the weights during training.]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Ryan’s system with Crotinger’s system, with a motivation for pattern detection, and more particularly relates to pattern detection in time-series data. Conventional systems may be used to detect patterns, but typically do not detect patterns in data obtained specifically from a time-series. Pattern detection in a time-series, according to the present disclosure, may be used in the field of Machine Learning (ML) for networking applications, telecommunications, as well as many other applications. [Ryan, para. 61]
Regarding claim 14, modified Crotinger teaches the method of claim 1, but Crotinger does not teach wherein the plurality of time series comprise one or more multi-dimensional time series, and obtaining the plurality of time series comprises aggregating at least two obtained single-dimensional time series to form each multi-dimensional time series.
However, Ryan does teach wherein the plurality of time series comprise one or more multi-dimensional time series, and obtaining the plurality of time series comprises aggregating at least two obtained single-dimensional time series to form each multi-dimensional time series. [Ryan, para. 13 discloses The one-dimensional or multi-dimensional windows are created either independently or jointly with the time-series. The method also includes training a Deep Neural Network (DNN) with the one-dimensional or multi-dimensional windows utilizing historical and/or simulated data to provide a neural network model. The method further includes processing ongoing data from a network with the neural network model to detect one or more patterns of a particular category in the ongoing data and localizing the one or more patterns in time.]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Ryan’s system with Crotinger’s system, with a motivation for pattern detection, and more particularly relates to pattern detection in time-series data. Conventional systems may be used to detect patterns, but typically do not detect patterns in data obtained specifically from a time-series. Pattern detection in a time-series, according to the present disclosure, may be used in the field of Machine Learning (ML) for networking applications, telecommunications, as well as many other applications. [Ryan, para. 61]
Regarding claim 15, modified Crotinger teaches the method of claim 1, but Crotinger does not teach wherein the number of models of non-anomalous network behavior is less than the number of time series.
However, Ryan does teach wherein the number of models of non-anomalous network behavior is less than the number of time series. [Ryan, para. 127 discloses Time-series data may principally correlate to human behavior, machine dynamics, or any other system that exhibits cyclic patterns. In applications that produce periodic temporal data, effective detection of anomalies generally includes complex, non-trivial, error-prone, and empirical processes. The optimal anomaly detection strategy thus depends on proactive model selection and hyper-parameter optimization by employing meta-learning. Furthermore, deep learning models (i.e., DNNs) create abstract representations that overcome the processing and communication limits and generalize to a variety of data types.]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Ryan’s system with Crotinger’s system, with a motivation for pattern detection, and more particularly relates to pattern detection in time-series data. Conventional systems may be used to detect patterns, but typically do not detect patterns in data obtained specifically from a time-series. Pattern detection in a time-series, according to the present disclosure, may be used in the field of Machine Learning (ML) for networking applications, telecommunications, as well as many other applications. [Ryan, para. 61]
Regarding claim 16, modified Crotinger teaches the method of claim 1, but Crotinger does not teach wherein the plurality of time series represent a corresponding plurality of marginal distributions of performance of the multi-domain communication system.
However, Ryan does teach wherein the plurality of time series represent a corresponding plurality of marginal distributions of performance of the multi-domain communication system. [Ryan, para. 75 discloses (1) the time-series that contains the pattern and (2) the indicator to be associated with the pattern. A time-series is used to define training instances using the windowing approach, defined in more detail below, while the indicator is used to associate a class with the instance. Due to the classification capacity of Deep Neural Networks (DNNs), it is not necessary to be precise with selection of the duration of the time-series. With sufficient training, the network can self-adjust to find the pattern. In the example of pattern detection in SNR analysis, the time-series included measurements and the indicator was the threshold crossing. Notice that the indicator can be something completely different from the time-series, such as the loss of a video signal, when the time-series relates to the fill level of network buffers.]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Ryan’s system with Crotinger’s system, with a motivation for pattern detection, and more particularly relates to pattern detection in time-series data. Conventional systems may be used to detect patterns, but typically do not detect patterns in data obtained specifically from a time-series. Pattern detection in a time-series, according to the present disclosure, may be used in the field of Machine Learning (ML) for networking applications, telecommunications, as well as many other applications. [Ryan, para. 61]
Regarding claim 20, it recites features similar to features within claim 1, therefore, it is rejected in a similar manner.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over US 20180324199 A1 to Crotinger et al., (hereinafter, “Crotinger”) in view of US 20210089927 A9 to Ryan et al., (hereinafter, “Ryan”) in further view of US 20220342860 A1 to Gonzalez Macias et al., (hereinafter, “Gonzalez”).
Regarding claim 6, modified Crotinger teaches the method of claim 5, but modified Crotinger does not teach wherein one or more of the following applies: detecting whether each of the time series includes a seasonal component is based on one of the following statistical tests: a Welch test, or a QS test; and detecting whether each of the time series includes a non-constant trend component is based on one of the following: a stationarity test, a Kolmogorov-Smirnov test, or a neural network autoencoder.
However, Gonzalez does teach wherein one or more of the following applies: detecting whether each of the time series includes a seasonal component is based on one of the following statistical tests: a Welch test, or a QS test; [Gonzalez, para. 65 discloses model matching subsystem 116 may adjust thresholds for different statistical tests/models such as a p-value for a seasonal test, a p-value for a serial-correlated test based on the timeseries signal. In some embodiments, model matching subsystem 116 may adjust general probability thresholds based on the timeseries signal. One or more of those parameters may be model execution parameters.] and detecting whether each of the time series includes a non-constant trend component is based on one of the following: a stationarity test, a Kolmogorov-Smirnov test, or a neural network autoencoder. [Gonzalez, para. 59 discloses temporal trait detection subsystem 114 may execute a combination of Dickey-Fuller and Kwiatkowski-Phillips-Schmidt-Shin (KPSS) for stationary and trend-stationary tests to determine whether the series data exhibits a linear trend (e.g., whether the mean changes linearly). As a result of executing these tests, temporal trait detection subsystem 114 may determine whether the series data has trend.]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Gonzalez’s system with modified Crotinger’s system, with a motivation for temporal trait detection subsystem 114 may determine a temporal trait associated with the timeseries dataset. As referred to herein, the temporal trait identifies a pattern within the timeseries dataset. Temporal trait detection subsystem 114 may execute a function for determining a temporal trait of a timeseries dataset. The temporal traits may include a determination whether the data has a trend, heteroskedasticity, seasonality, an approximate constant, or another suitable temporal trait. [Gonzalez, para. 56]
Claims 17 – 19 are rejected under 35 U.S.C. 103 as being unpatentable over US 20180324199 A1 to Crotinger et al., (hereinafter, “Crotinger”) in view of US 20210089927 A9 to Ryan et al., (hereinafter, “Ryan”) in further view of US 20210306372 A1 to Koral et al., (hereinafter, “Koral”).
Regarding claim 17, modified Crotinger teaches the method of claim 1, but modified Crotinger does not teach wherein: the multiple domains include at least two of the following domains: a user equipment, UE, domain; a radio access network, RAN, domain; a core network, CN, domain; and an IP multimedia system, IMS, domain; and the plurality of time series include at least one time series obtained from each of the at least two domains.
However, Koral does teach wherein: the multiple domains include at least two of the following domains: a user equipment, UE, domain; [Koral, para. 40 discloses the system 200 may include core network 12, customer premise 30, an enterprise MEC 205, 5G RAN 34, user equipment (UE) 210, and UE 212. The UE 210 and UE 212 may be one or more IoT devices 32 and one or more mobile devices 33, e.g., cellular phones, satellite communication systems, portable digital assistants (PDAs), laptop computers, tablet devices, smart meters, commercial security systems, traffic monitoring systems, weather monitoring systems, vehicles, smart home devices and sensors, and so on.] a radio access network, RAN, domain; [Koral, para. 37 discloses That U-PGW 36 may thus provide a gateway to and from the Enterprise WAN 44 from and to the 5G Radio Access Network (RAN) 34. The 5G RAN 34 may include one or more base stations and operate in conjunction with a RAN Intelligent Controller (MC), illustrated in FIG. 2. The RIC may include a set of functions and interfaces that allow for increased optimizations through policy-driven closed loop automation for faster, more flexible service deployments and programmability within the 5G RAN 34. One or more IoT devices 32 and one or more mobile devices 33 may also access the 5G RAN 34.] a core network, CN, domain; [Koral, para. 74 discloses Applications can include, without limitation, web browsing, VoIP, streaming media, and the like. Each application can pose different Quality of Service (QoS) requirements on a respective packet transfer. Different packet transfers can be served by different bearers within core network 404, e.g., according to parameters, such as the QoS.] and an IP multimedia system, IMS, domain; [Koral, para. 72 discloses External networks 406 can include one or more trusted and non-trusted external networks such as an internet protocol (IP) network 408, an IP multimedia subsystem (IMS) network 410, and other networks 412, such as a service network, a corporate network, or the like.] and the plurality of time series include at least one time series obtained from each of the at least two domains. [Koral, para. 53 discloses the security component 220 may monitor signal power measurements of mobile devices 33 connected to 5G RAN 34. Signal power measurements may be associated with a signal power between a mobile device and a cell tower, which may roughly indicate a distance between the mobile device and cell tower. The anomaly detection element 221 may apply analytics (e.g., a time series anomaly detection method that tracks Reference Signals Received Power (RSRP) (power measurements) from eNBs/gNBs in a region and detects eNBs/gNBs that are active for short time periods) in order to detect abnormalities in operations for the mobile device 33 (e.g., entities id, location, and operation patterns).]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Koral’s system with modified Crotinger’s system, with a motivation that by running applications and performing related processing tasks closer to the cellular customer, core network data traffic is reduced and applications respond faster due to the reduced physical distance between client and server, and energy can be saved eventually by offloading processing into the cloud despite adding additional communication overhead and latency to the communication. [Koral, para. 22]
Regarding claim 18, modified Crotinger teaches the method of claim 17, but modified Crotinger does not teach wherein: the RAN domain comprises an Open RAN, O-RAN, architecture; the obtaining, determining, and classifying operations are performed by an O-RAN non-real-time RAN intelligent controller, non-RT RIC; and the detecting operation is performed by the O-RAN non-RT RIC or by an O-RAN near-RT RIC.
However, Koral does teach wherein: the RAN domain comprises an Open RAN, O-RAN, architecture; [Koral, para. 37 discloses That U-PGW 36 may thus provide a gateway to and from the Enterprise WAN 44 from and to the 5G Radio Access Network (RAN) 34. The 5G RAN 34 may include one or more base stations and operate in conjunction with a RAN Intelligent Controller (MC), illustrated in FIG. 2. The RIC may include a set of functions and interfaces that allow for increased optimizations through policy-driven closed loop automation for faster, more flexible service deployments and programmability within the 5G RAN 34. One or more IoT devices 32 and one or more mobile devices 33 may also access the 5G RAN 34.] the obtaining, determining, and classifying operations are performed by an O-RAN non-real-time RAN intelligent controller, non-RT RIC; [Koral, para. 37 discloses The RIC may include a set of functions and interfaces that allow for increased optimizations through policy-driven closed loop automation for faster, more flexible service deployments and programmability within the 5G RAN 34. One or more IoT devices 32 and one or more mobile devices 33 may also access the 5G RAN 34.] and the detecting operation is performed by the O-RAN non-RT RIC or by an O-RAN near-RT RIC. [Koral, para. 43 discloses The anomaly detection element 221 may collect RAN data via the RIC 215 in response to network traffic generated by each of the UEs. The anomaly detection element 221 may monitor the collected RAN data for patterns. For example, the anomaly detection element 221 may monitor trends in control messages (e.g., radio resource control (RRC)) within the RAN data, as well as performance telemetry data. The anomaly detection element 221 may analyze the collected RAN data to detect surges on the enterprise network 20, which may include a large amount of UEs (e.g., greater than 30% of all UEs or greater than 30% of a subset of UEs of a particular UE type) connecting to the enterprise network 20 within a predetermined period (e.g., one minute, 5 minutes, etc.).]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Koral’s system with modified Crotinger’s system, with a motivation that by running applications and performing related processing tasks closer to the cellular customer, core network data traffic is reduced and applications respond faster due to the reduced physical distance between client and server, and energy can be saved eventually by offloading processing into the cloud despite adding additional communication overhead and latency to the communication. [Koral, para. 22]
Regarding claim 19, modified Crotinger teaches the method of claim 17, but modified Crotinger does not teach wherein the plurality of time series include at least two of the following: time series of one or more of the following RAN-domain quality of service, QoS, metrics: RAN resources used, serving cell load, mobility events between serving cells, and serving and neighbor cell radio measurements; time series of one or more of the following CN-domain QoS metrics: packet delay, packet delay jitter, packet loss, and priority level; time series of trace data for respective cells provided by RAN nodes; time series of performance management, PM, counter values associated with the RAN nodes; time series of user plane, UP, event information associated with the CN domain; and time series of control plane, CP, event information associated with the CN domain.
However, Koral does teach wherein the plurality of time series include at least two of the following: time series of one or more of the following RAN-domain quality of service, QoS, metrics: RAN resources used, serving cell load, mobility events between serving cells, and serving and neighbor cell radio measurements; [Koral, para. 43 discloses the anomaly detection element 221 may collect RAN data via the RIC 215 in response to network traffic generated by each of the UEs. The anomaly detection element 221 may monitor the collected RAN data for patterns. For example, the anomaly detection element 221 may monitor trends in control messages (e.g., radio resource control (RRC)) within the RAN data, as well as performance telemetry data. The anomaly detection element 221 may analyze the collected RAN data to detect surges on the enterprise network 20, which may include a large amount of UEs (e.g., greater than 30% of all UEs or greater than 30% of a subset of UEs of a particular UE type) connecting to the enterprise network 20 within a predetermined period (e.g., one minute, 5 minutes, etc.).] time series of one or more of the following CN-domain QoS metrics: packet delay, packet delay jitter, packet loss, and priority level; [Koral, para. 75 discloses Core network 404 uses a concept of bearers, e.g., EPS bearers, to route packets, e.g., IP traffic, between a particular gateway in core network 404 and UE 414. A bearer refers generally to an IP packet flow with a defined QoS between the particular gateway and UE 414. Access network 402, e.g., E UTRAN, and core network 404 together set up and release bearers as required by the various applications. Bearers can be classified in at least two different categories: (i) minimum guaranteed bit rate bearers, e.g., for applications, such as VoIP; and (ii) non-guaranteed bit rate bearers that do not require guarantee bit rate, e.g., for applications, such as web browsing.] time series of trace data for respective cells provided by RAN nodes; [Koral, para. 106 discloses RAN 604 may include one or more base stations 616, along with other network elements (not shown), such as a base station controller (BSC), a radio network controller (RNC), or relay nodes. One or more base stations 616 may be configured to transmit or receive wireless signals within a particular geographic region, which may be referred to as a cell (not shown). The cell may further be divided into cell sectors. For example, the cell associated with base station 616 may be divided into three sectors such that base station 616 may include three transceivers: one for each sector of the cell. In another example, base station 616 may employ multiple-input multiple-output (MIMO) technology and, therefore, may utilize multiple transceivers for each sector of the cell.] time series of performance management, PM, counter values associated with the RAN nodes; [Koral, para. 58 discloses a security component 220 within MEC 205, may monitor attachment rates of mobile devices 33 connecting to 5G RAN 34, detachment rates of mobile devices 33 disconnecting from 5G RAN 34 and signal power measurements of mobile devices 33 connected to 5G RAN 34. At block 272, the security component 220 may periodically query a mobile device 33 to determine whether the mobile device is still connected to the 5G RAN 34 or whether the mobile device 33 has reattached to the 5G RAN 34 in an abnormal manner (e.g., the mobile device 33 reattached at a rate that is three standard deviations or more from a measured network wide attach rate).] time series of user plane, UP, event information associated with the CN domain; [Koral, para. 112 discloses core network 606 may provide call control, billing services, mobile location-based services, pre-paid calling, Internet connectivity, video distribution or high-level security functions, such as user authentication.] and time series of control plane, CP, event information associated with the CN domain. [Koral, para. 113 discloses Core network 606 may also serve as a gateway for WTRUs 602 to access PSTN 608, Internet 610, or other networks 612. PSTN 608 may include circuit-switched telephone networks that provide plain old telephone service (POTS). For LTE core networks, core network 606 may use IMS core 614 to provide access to PSTN 608. Internet 610 may include a global system of interconnected computer networks or devices that use common communication protocols, such as the transmission control protocol (TCP), user datagram protocol (UDP), or IP in the TCP/IP internet protocol suite.]
Therefore, it would have been obvious to one of ordinary skill within the effective filling date to combine Koral’s system with modified Crotinger’s system, with a motivation that by running applications and performing related processing tasks closer to the cellular customer, core network data traffic is reduced and applications respond faster due to the reduced physical distance between client and server, and energy can be saved eventually by offloading processing into the cloud despite adding additional communication overhead and latency to the communication. [Koral, para. 22]
Conclusion
Pertinent prior art made of record however not relied upon:
US 20140278335 A1 to Macaro et al.
“Techniques for automated Bayesian posterior sampling using Markov Chain Monte Carlo and related schemes are described. In an embodiment, one or more values in a stationarity phase for a system configured for Bayesian sampling may be initialized. Sampling may be performed in the stationarity phase based upon the one or more values to generate a plurality of samples. The plurality of samples may be evaluated based upon one or more stationarity criteria. The stationarity phase may be exited when the plurality of samples meets the one or more stationarity criteria. Other embodiments are described and claimed.”
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Phuc Pham whose telephone number is (571)272-8893. The examiner can normally be reached Monday - Thursday 7:30 AM - 4:30 PM; Friday 8:00 AM - 12:00 PM.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Linglan Edwards can be reached at (571) 270-5440. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/P.P./Patent Examiner, Art Unit 2408
/TECHANE GERGISO/Primary Examiner, Art Unit 2408
